CN1362789A - Turbo code decoding method and decoder - Google Patents

Abstract

The invention discloses a turbo code decoding method and a decoder thereof. On the basis of the maximum a posteriori probability decoding method, the method adds the calculation and control of the state metrics at the initial several moments through the state grid graph; the decoding The α and β initial state measure calculation control modules are added in the controller to control the former state measure and the initial subsequent state measure to eliminate the introduction of redundant state measure. The present invention has a performance gain of about 0.1db compared to the decoding performance improvement without initial state metric calculation control.

Description

Turbine Code Decoding Method and Its Decoder

The present invention relates to channel coding and decoding technology in the field of mobile communication, more specifically to a turbo code decoding method and a decoder thereof which can improve decoding performance.

In a wireless communication system, due to the inhomogeneity and instability of the transmission medium, the transmitted signal will be affected by interference such as time diffusion and fading, resulting in random errors in the received bits. In order to prevent the interference effect of channel noise, it is necessary to adopt a certain method to improve the reliability and effectiveness of information transmission. The error correction coding method, which reduces the bit error rate by increasing redundancy, has been proved to be an effective and reliable important means by time. Especially in mobile communication and satellite communication systems. Error-correcting codes are widely used in technology.

In 1993, Berrou et al. proposed a Turbo (Turbo, the same below) code, which is a decodable code close to a pseudo-random code. Tebo is a regression systematic convolutional code (RSC). The superiority of this coding method comes from its unique code structure and iterative decoding algorithm. Iterative decoding breaks through the design idea of minimum code distance, and is closer to the concept of Shannon random code.

Turbine code has been proved to be a code with strong error correction ability. Its encoder is composed of two or more sub-encoders connected in series or in parallel. Usually, the more common turbo code encoder is composed of two convolutional code encoders in parallel, and the input information bits are all the way It is directly sent to the sub-encoder 1, and the other channel is sent to the sub-encoder 2 for encoding after passing through the interleaver. After encoding, the encoded data is then punctured by a puncher and modulated to a suitable code rate for output. As shown in Figure 1.

Figure 1 shows the structure of the turbo code encoder in the cdma2000 and WCDMA proposals. The function of the interleaver is to rearrange the order of the input data, and the purpose is to adjust the weight distribution so that the weight distribution of the sub-encoder 2 input bit stream is consistent with Sub-encoder 1 is different. The puncturer performs puncture sampling and parallel-to-serial conversion on the six bits output by the two sub-encoders.

A convolutional code encoder usually uses (n0, k0, m) to represent features, where n0 is the output bit, k0 is the input bit, and m is the number of registers. K represents the constraint length, that is, the number m of shift registers in the convolutional code plus 1. In the cdma2000 proposal, the convolutional code encoder of the turbo code sub-(3, 1, 4) regression system is shown in Fig. 2 .

What Fig. 2 shows is a R = 1/3 code rate regression systematic convolution code encoder (RSC). 20 is a shift register, and there are three shift registers in total, so m=3, K=4. 22 is a modulo 2 adder. 24 is the tail bit control structure. After a frame of data input is completed, the 20 register needs to be cleared. This is to switch the tail bit controller switch to the bottom. Through three beats, the bits in the three registers 20 are used as input in sequence cleared.

The RSC sub-encoder of the turbo code encoder in the wideband code division multiple access (WCDMA) proposal is similar to that in Figure 2, except that there is no output terminal of Y1, that is, the regression system convolutional code encoding of (2, 1, 3) controller (RSC).

The decoder of the Turbo code adopts an iterative and recursive method to continuously improve the decoding accuracy through multiple iterations. Fig. 3 is a structure of a turbo code decoder, in which 33 and 34 refer to a soft-input soft-output (SISO) decoder. 31 is a depuncturing device, corresponding to the inverse operation of the puncturer in the encoder, 32 is a deinterleaver, corresponding to the inverse operation of the interleaver in the encoder, and restores the order before interleaving, 35 is a symbol decision device, when the input When the data is greater than or equal to 0, output 1; when the input data is less than 0, output 0.

Soft-Input Soft-Output (SISO) decoders are mainly divided into two methods: maximum a posteriori probability decoding (MAP) and maximum likelihood decoding (SOVA).

The Maximum A Posteriori Probability Decoding (MAP) method is based on the regressive convolutional coding trellis graph, and obtains the Maximum Posteriori Probability Log Ratio (LLR) of each decoded bit through forward and backward recursion.

d _k ＝1 if^(d _k )＞0 (2)

d _k =0 if^(d _k )≤0 where d _k corresponds to encoded input data; R ₁ ^N is the N sets of decoded data received by the decoder. import definition: $a_k^i\left(m\right)=\frac{P_r(d_k=i,S_k=m,R_1^k)}{P_r\left(R_1^k\right)}=P_r(d_k=i,S_k=m/R_1^k)----\left(3\right)$ ${\mathrm{\β}}_k\left(m\right)=\frac{P_r(R_{k+1}^N/S_k=m)}{P_r(R_{k+1}^N/R_1^N)}---\left(4\right)$

γ _i (R _k , m', m)=P _r (d _k =i, R _k , S _k =m/S _k-1 =m') (5)

α和β满足如下递推关系 $a_k^i\left(m\right)=\frac{\underset{m^{\′}}{\mathrm{\Σ}}\underset{j=0}{\overset{1}{\mathrm{\Σ}}}{\mathrm{\γ}}_i(R_k,m^{\′},m)a_{k-1}^j\left(m^{\′}\right)}{\underset{m}{\mathrm{\Σ}}\underset{m^{\′}}{\mathrm{\Σ}}\underset{i=0}{\overset{1}{\mathrm{\Σ}}}\underset{j=0}{\overset{1}{\mathrm{\Σ}}}{\mathrm{\γ}}_i(R_k,m^{\′},m)a_{k-1}^j\left(m^{\′}\right)}---\left(7\right)$ ${\mathrm{\β}}_k\left(m\right)=\frac{\underset{m^{\′}}{\mathrm{\Σ}}\underset{j=0}{\overset{1}{\mathrm{\Σ}}}{\mathrm{\γ}}_i(R_{k+1},m^{\′},m){\mathrm{\β}}_{k+1}\left(m^{\′}\right)}{\underset{m}{\mathrm{\Σ}}\underset{m^{\′}}{\mathrm{\Σ}}\underset{i=0}{\overset{1}{\mathrm{\Σ}}}\underset{j=0}{\overset{1}{\mathrm{\Σ}}}{\mathrm{\γ}}_i(R_{k+1},m^{\′},m)a_k^j\left(m^{\′}\right)}---\left(8\right)$ 根据MAP解码理论，最大后验概率解码(MAP)方法由以下几步完成：步骤1：初始化状态度量α、β $a_0^i\left(0\right)=1{,a}_0^i\left(m\right)=0--\∀m\≠0--i=\mathrm{0,1}----\left(9\right)$ Where: S _k =m represents the mth state at the Kth moment; α, β, and γ correspond to the previous state metric, reverse state metric, and branch metric in the state transition grid diagram, respectively. At this time, the logarithmic likelihood ratio (LLR) of the decoding posterior probability can be expressed as:

α and β satisfy the following recurrence relation $a_k^i\left(m\right)=\frac{\underset{m^{\′}}{\mathrm{\Σ}}\underset{j=0}{\overset{1}{\mathrm{\Σ}}}{\mathrm{\γ}}_i(R_k,m^{\′},m)a_{k-1}^j\left(m^{\′}\right)}{\underset{m}{\mathrm{\Σ}}\underset{m^{\′}}{\mathrm{\Σ}}\underset{i=0}{\overset{1}{\mathrm{\Σ}}}\underset{j=0}{\overset{1}{\mathrm{\Σ}}}{\mathrm{\γ}}_i(R_k,m^{\′},m)a_{k-1}^j\left(m^{\′}\right)}---\left(7\right)$ ${\mathrm{\β}}_k\left(m\right)=\frac{\underset{m^{\′}}{\mathrm{\Σ}}\underset{j=0}{\overset{1}{\mathrm{\Σ}}}{\mathrm{\γ}}_i(R_{k+1},m^{\′},m){\mathrm{\β}}_{k+1}\left(m^{\′}\right)}{\underset{m}{\mathrm{\Σ}}\underset{m^{\′}}{\mathrm{\Σ}}\underset{i=0}{\overset{1}{\mathrm{\Σ}}}\underset{j=0}{\overset{1}{\mathrm{\Σ}}}{\mathrm{\γ}}_i(R_{k+1},m^{\′},m)a_k^j\left(m^{\′}\right)}---\left(8\right)$ According to the MAP decoding theory, the maximum a posteriori probability decoding (MAP) method is completed by the following steps: Step 1: Initialize the state metrics α, β $a_0^i\left(0\right)=1{,a}_0^i\left(m\right)=0--\∀m\≠0--i=\mathrm{0,1}----\left(9\right)$

β _N (0)=1, β _N (m)=0 m≠0(10)

Step 2: For each symbol received, calculate the branch metric γ and the previous state metric α

Step 3: After the calculation of α of the received sequence is completed, calculate the subsequent state measure β, and obtain the logarithmic likelihood ratio LLR of the posterior probability of each bit,

Usually, the MAP decoding method is mainly carried out according to the above three steps, and the whole decoding process is mainly the calculation of branch metric γ, state metric α, β and LLR.

Since there are many exponential and multiplication operations in standard MAP decoding, the implementation of MAP decoding is often performed in the logarithmic domain, which is Log MAP. In this way, the multiplication operation in the standard MAP decoding becomes addition, and the addition operation can be simplified by the Jacobian formula and realized by a look-up table. The Jacobian formula is as follows:

l _n (e ^x +e ^y )=max(x,y)+ln(1-exp{-|xy|})max(x,y)+f _c (|xy|) (11)

Moving from Standard MAP to Log MAP is algorithmically equivalent. In the processing of state metrics at the initial moment, the usual method is to increase the weight of the initial zero state metric, so that the state metric value of the state transition from the initial zero state is larger than other state metric values. In this way, in the implementation process of Log MAP decoding, it is only necessary to transfer the implementation of the entire decoder to the logarithmic domain according to the implementation steps of MAP decoding.

According to the steps of the MAP decoding method, in the process of MAP decoding, the initialization of the previous state metric α and the subsequent state metric must be performed first, wherein the state metrics are initialized according to formulas (9) and (10), because in the standard MAP decoding In , on the state transition path, the state metric at the current moment is multiplied by the state metric at the previous moment and the branch metric, so the initialization according to (9) and (10) will not affect the decoding result. When the standard MAP decoding is transferred to the logarithmic domain, the multiplication relationship in the state metric calculation in the standard MAP decoding becomes an additive relationship. At this time, no matter what kind of initialization is performed, redundant The state transition path of , as shown by the dotted line in Figure 4.

In Fig. 4, Fig. 4 is the grid diagram of the initial state transition of the previous state metric, and Fig. 5 is the initial state transition grid of the subsequent state metric; 43 is the state transition path; 44 represents a certain state of the encoder at a certain moment.

When dealing with the initial state metrics, the usual method is to take a larger value for the state metrics of the zero state, and take a relatively small value for the state metrics of other states, so that the state metrics transferred from the zero state can be as far as possible The value takes a larger weight. However, even if the method of increasing the initial zero-state state metric is adopted, the calculation of the subsequent state metric by the state metric of the additional transition path cannot be completely eliminated, which is equivalent to introducing artificial noise in the calculation of the state metric, thereby affecting decoding. performance.

For this reason, the purpose of the present invention is to address the shortcomings of the above-mentioned Turbo code decoding method that cannot eliminate the calculation of the additional path state metric to the subsequent stage state metric, which introduces artificial noise and affects the decoding performance, and proposes a method that belongs to the maximum a posteriori probability A turbo code decoding method and a decoder thereof.

In order to achieve the above object, the present invention adopts the following technical solutions:

The turbo code decoding method is based on the maximum a posteriori probability decoding method, and first initializes the antecedent state metric α and the reverse state metric β of the input decoded soft data information, and the method also includes the following steps:

a, Calculate and control the state metrics at the initial few moments through the state grid diagram to eliminate the introduction of redundant state metrics;

b. Calculate the branch metric γ and the previous state metric α for each received symbol of the decoded soft data information;

c. After the calculation of the previous item state metric α of the above-mentioned symbol is completed, the subsequent item state metric β is calculated, and the posterior probability logarithmic logic likelihood ratio LLR of each bit is obtained;

d, After multiple iterations of the above steps, output the soft output decoding information.

A Tebow code decoder, the decoder is based on the maximum a posteriori probability state transition trellis graph decoder, which includes a branch metric gamma calculation module for calculating the branch metric value in the state transition path, for calculating the state transition The former state measure α calculation module and memory of the state measure value in the state transition, the subsequent state measure β calculation module and memory used to calculate the state measure value in the state transition, the former state measure α and the latter state measure β Initialized state metric initialization module, lookup table, logic likelihood ratio calculation module, general control module that generates soft input and soft output decoding control signals, branch metric γ calculation module receives decoding input data, and its output is connected to the previous state metric α The calculation module and the subsequent state measure β calculation module, the calculation results of the previous state measure α calculation module and the subsequent state measure β calculation module are stored in the α and β memories respectively, and the lookup table provides the former state measure α calculation module, the latter The state metric β calculation module and the logic likelihood ratio calculation module look up the table, and the α and β memories are respectively output to the logic likelihood ratio calculation module. The decoder also includes an α initial state metric calculation control module and a β initial state metric calculation control module. The α initial state metric calculation control module and the state metric initialization module are bidirectionally connected, and the α initial state metric calculation control module and the β initial state metric calculation control module perform state measurement on the previous state metric α calculation module and the subsequent state metric β calculation module respectively. For transfer path control, the general control module provides control signals for the entire decoder, and the signals are respectively connected to the γ calculation module, the α initial state metric calculation control module and the α calculation module, the β initial state metric calculation control module, and the β calculation module.

Because the method of the present invention is on the basis of the maximum a posteriori probability decoding method, it increases the calculation and control of the state metrics at the initial few moments through the state grid diagram, eliminating the introduction of redundant state metrics; and according to the method of the present invention In the designed decoder, the α initial state metric calculation control module and the β initial state metric calculation control module are added, and the state transition path control is performed on the previous state metric and the initial subsequent state metric to eliminate the introduction of redundant state metrics, which can improve Accuracy of state metrics and improved decoding performance. Therefore, the decoding performance of the Log MAP after the initial state metric calculation control is improved by about 0.1db compared to the decoding performance without the initial state metric calculation control. That is to say, in the case of ensuring the same bit error rate, the initial state metric calculation control processing decoder can tolerate worse input signal quality (the input signal signal-to-noise ratio is smaller), and in wireless communication, the emission can be reduced Machine power, reduce interference, improve communication quality.

Below in conjunction with accompanying drawing and embodiment, the decoding method of the present invention is described in detail:

FIG. 1 is a schematic diagram of the structure and principle of a traditional turbo code encoder.

FIG. 2 is a schematic diagram of the structural principle of a sub-encoder of the encoder in FIG. 1 .

FIG. 3 is a schematic diagram of the structure and principle of a traditional decoder corresponding to the encoder in FIG. 1 .

FIG. 4 is a grid diagram of an initial state transition of antecedent state metrics in an existing decoding method.

Fig. 5 is a grid diagram of the initial state transition of the latter state metric in the existing decoding method.

Fig. 6 is a schematic diagram of the structural principle of the decoder of the present invention.

In order to improve the decoding performance as much as possible, the introduction of redundant state transition paths must be eliminated during the Log MAP decoding process. In the standard MAP theory, in the process of calculating the previous state metric α and the subsequent state metric β, the branch metric γ and the previous state metric α are multiplicative coefficients, so when calculating the state metric, only need to The initial moment (the former state metric α corresponds to the first moment, and the latter state metric β corresponds to the last moment) is initialized according to the above formulas (9) and (10), so that the state metric initialized to zero and the corresponding state transition metric γ The product of is still zero, so it should not introduce redundant state metrics to affect the calculation of subsequent state metrics. However, in the Log MAP operation, in order to reduce the amount of MAP decoding operations, all operations are transferred to the logarithmic domain. In this way, the original multiplication becomes the addition of the logarithmic domain. At this time, if the state metric is initialized according to the formulas (9) and (10), since the state transition metric is not zero, when calculating the next state metric, it will be A redundant state metric value will be introduced, and this redundant state metric value will affect the calculation accuracy of the subsequent state metric value, thus affecting the decoding performance.

Based on the above considerations, the method of the present invention controls the calculation of the state metrics at the initial few moments through the state transition grid diagram to eliminate the introduction of redundant state metrics. The specific decoding method is still based on the maximum a posteriori probability decoding method. First, the input The previous item state measure α and the reverse state measure β of the decoded soft data information are initialized, and the method also includes the following steps:

a, Calculate and control the state metrics at the initial few moments through the state grid diagram to eliminate the introduction of redundant state metrics;

b. Calculate the branch metric γ and the previous state metric α for each received symbol of the decoded soft data information;

c. After the calculation of the previous item state metric α of the above-mentioned symbol is completed, the subsequent item state metric β is calculated, and the posterior probability logarithmic logic likelihood ratio LLR of each bit is obtained;

d. After multiple iterations of the above steps, the soft output decoding information is finally output.

Both the branch metric γ and the previous state metric α in the step b are transition probabilities, the branch metric γ is determined by the channel model and the state transition diagram, and the previous state metric α is calculated by formula (7).

In the described step c, after the calculation of the previous item state measure α of the symbol is completed, the subsequent item state measure of the symbol is calculated according to formula (8), and when calculating the latter item state measure β, it corresponds to each symbol The logarithmic likelihood ratio LLR of the posterior probability is calculated.

In said step d, said multiple iterations means that the LLR output of each iterative decoding is deinterleaved and used as the input of the next iterative decoding, and after repeated cycles, the decoded information is finally output.

In the above-mentioned method of the present invention, the so-called state measurement meaning of several initial moment is, according to the state transition grid diagram (shown in conjunction with Fig. 4, 5), the initial moment at the former state measurement and the subsequent state measurement is not All states undergo state transition, so it is necessary to control the calculation of the initial state metric according to the actual state transition. As described in the background technology, when processing the state metric at the initial moment, the usual method is to take a zero state metric and take one A larger value, while the other state metrics take a relatively small value to try to make the state metrics transitioning from the zero state account for a larger weight.

Since not all states undergo state transition at the initial moment of state metric calculation, if the initial calculation is not controlled, state metric calculation will be performed for all states, which will introduce redundant state metrics, which will affect Subsequent state metric calculations have an impact, thereby degrading decoding performance. Calculating and controlling the state metrics at the initial moment will eliminate the influence of redundant state metrics, thereby improving the decoding performance.

On the basis of the soft-input-output (SISO) decoder of the traditional Turbo code decoder, according to the above-mentioned method of the present invention, the decoder is based on the state transition trellis graph decoder of the maximum a posteriori probability, and it still includes for calculating The branch metric gamma calculation module of the branch metric value in the state transition path, the previous state metric α calculation module and memory for calculating the state metric value in the state transition, and the subsequent item state for calculating the state metric value in the state transition Metric β calculation module and memory, a state metric initialization module for initializing the previous state metric α and the subsequent state metric β, a lookup table, a logic likelihood ratio calculation module, and a total control module for generating control signals for soft input and soft output decoding , the branch metric γ calculation module receives the decoded input data, and its output is connected to the previous state metric α calculation module and the subsequent state metric β calculation module, and the calculation results of the previous state metric α calculation module and the subsequent state metric β calculation module are respectively Stored in the α and β memories, the lookup table provides the previous state measure α calculation module, the subsequent state measure β calculation module, and the logical likelihood ratio calculation module look-up table, and the α and β memories are respectively output to the logical likelihood ratio calculation module. The decoder also includes an α initial state metric calculation control module and a β initial state metric calculation control module, the α initial state metric calculation control module is bidirectionally connected with the state metric initialization module, the α initial state metric calculation control module and the β initial state metric calculation The control module performs state transition path control on the previous state metric α calculation module and the subsequent state metric β calculation module, and the general control module provides the control signal of the entire decoder, which is respectively connected to the branch metric γ calculation module, α initial state The metric calculation control module and the previous state metric α calculation module, the β initial state metric calculation control module and the subsequent state metric β calculation module.

Please in conjunction with shown in Fig. 6, in this figure, unit 51 is a branch metric γ (Gamma) calculation module, is used for calculating the branch metric value in the state transfer path; Unit 52 is a preceding item state metric α (Alpha) calculation module, For calculating the previous item state metric value in the state transition; Unit 53 is the latter item state metric β (Beta) calculation module, for calculating the latter item state metric value in the state transition; Unit 54 is the logical likelihood ratio LLR calculation module , calculate the posterior probability logarithmic ratio according to the state metric value; unit 55 is a state metric initialization module, and completes the initialization of the former item state metric α and the subsequent item state metric β; unit 56 is an eight-valued lookup table; unit 57 is α The memory is used to store the previous item state metric value that is calculated; the unit 58 is a β memory, and is used to store the subsequent item state metric value that is calculated; the unit 59 is a total control module that generates soft input and soft output decoding control signals; the unit 5a is the calculation control module of the initial antecedent state metric α, so that the calculation of the antecedent state metric is completely carried out according to the actual state transition path (as shown by the solid lines in Figure 4 and 4b); unit 5b is the initial subsequent state metric β The calculation control module makes the calculation of the latter state measure completely follow the actual state transition path.

The method and the decoder of the present invention have undergone decoding fixed-point simulation, and the Log MAP decoding performance after initial control processing is improved by about 0.1dB performance gain compared with the decoding performance without initial control.

Simulation conditions:

Decoded code block length 5114

Decoding input data 6Bits quantization

When there is no initial state metric calculation control module, the initialization is as follows: $\ln \left[a_0^i\left(0\right)\right]=63,\ln \left[a_0^i\left(m\right)\right]=0--\∀m\≠0--i=\mathrm{0,1}---\left(12\right)$

ln[β _N (0)]=63, ln[β _N (m)]=0 m≠0 (13)

When there is an initial state metric calculation control module, the simulation performance is not affected by the initial value of the state metric. In the case of ensuring the same bit error rate, the initial state metric calculation control processing decoder can tolerate worse input signal quality (the input signal signal-to-noise ratio is smaller), and in wireless communication, the power of the transmitter can be reduced , Reduce interference and improve communication quality. In CDMA2000 and WCDMA systems, the transmit power of the base station can be reduced and the capacity of the cell can be increased.

Claims (5)

1. A turbo code decoding method, the decoding method is based on the method of state transition lattice diagram decoding of maximum a posteriori probability, at first to the preceding item state measure α, reverse state measure β of the decoding soft data information symbol of input Carry out initialization, it is characterized in that, this method also comprises the following steps: a, Calculate and control the state metrics at the initial few moments through the state grid diagram to eliminate the introduction of redundant state metrics; b. Calculate the branch metric γ and the previous state metric α for each received symbol of the decoded soft data information; c. After the calculation of the previous item state metric α of the above-mentioned symbol is completed, the subsequent item state metric β is calculated, and the posterior probability logarithmic logic likelihood ratio LLR of each bit is obtained; d, After multiple iterations of the above steps, output the soft output decoding information. 2. The turbo code decoding method according to claim 1, characterized in that: both the branch metric γ and the previous state metric α in the step b are transition probabilities, and the branch metric γ is determined by the channel model and the state transition diagram Decide. 3. The turbo code decoding method as claimed in claim 1, characterized in that: in the step c, when calculating the posterior state measure β, the logarithmic logarithm of the posterior probability corresponding to each symbol is Then the ratio LLR is calculated. 4. The turbo code decoding method according to claim 1, characterized in that: in the step d, the multiple iterations refer to the log likelihood ratio LLR output of each iteration decoding after deinterleaving As the input of the next iterative decoding, after repeated loops, the decoding information is finally output. 5. A turbo code decoder, the decoder is based on the state transition lattice graph decoder of maximum a posteriori probability, which includes a branch metric gamma calculation module for calculating the branch metric value in the state transition path, for calculating The previous state measure α calculation module and memory of the state measure value in the state transition, the subsequent state measure β calculation module and memory for calculating the state measure value in the state transition, and the former state measure α and the latter state measure β initialization state metric initialization module, lookup table, logic likelihood ratio calculation module, general control module that generates soft input and soft output decoding control signals, branch metric γ calculation module receives decoding input data, and its output is connected to the previous state The metric α calculation module and the subsequent state metric β calculation module, the calculation results of the previous state metric α calculation module and the subsequent state metric β calculation module are stored in α and β memories respectively, and the lookup table provides the previous state metric α calculation module, The subsequent state measure β calculation module and the logical likelihood ratio calculation module look up the table, and the α and β memories are respectively output to the logical likelihood ratio calculation module. It is characterized in that: the decoder also includes an α initial state metric calculation control module and a β initial The state metric calculation control module, the α initial state metric calculation control module and the state metric initialization module are bidirectionally connected, and the α initial state metric calculation control module and the β initial state metric calculation control module respectively calculate the previous state metric α calculation module and the subsequent state The metric β calculation module performs state transfer path control, and the general control module provides control signals for the entire decoder, which are respectively connected to the γ calculation module, the α initial state metric calculation control module, and the α calculation module, β initial state metric calculation control module and β calculation module.

Images